Research map / 3D vision / 3D generation
Motion Inspired Unsupervised Perception and Prediction in Autonomous Driving
Mahyar Najibi, Jingwei Ji, Yin Zhou, Charles R. Qi, Xinchen Yan, Scott Ettinger, Dragomir Anguelov
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| static point removal, scene decomposition and local consistency (NSFP++ over NSFP) | NSFP · WOD scene flow (val) | EPE3D (m): 0.455 -> 0.017 | Improved |
| NSFP++ over NSFP | NSFP · WOD scene flow (val) | mIoU over speed bins: 0.244 -> 0.586 | Improved |
| add box query with expansion (ablation) | NSFP++ · WOD scene flow ablation (val) | mIoU: 0.404 -> 0.552 | Mixed |
| add box query with expansion (ablation) | NSFP++ · WOD scene flow ablation (val) | EPE3D (m): 0.020 -> 0.023 | Hurt |
| add pruning to box query (ablation) | NSFP++ · WOD scene flow ablation (val) | EPE3D (m): 0.023 -> 0.018 | Improved |
| add local consistency loss (ablation) | NSFP++ · WOD scene flow ablation (val) | mIoU: 0.571 -> 0.586 | Improved |
| add AML pipeline to unsupervised flow (versus clustering only) (ablation) | Unsup Flow · WOD unsupervised moving-object 3D detection (PointPillars) | 3D mAP L1: 30.4 -> 42.1 | Improved |
| add scene flow to AML pipeline (ablation) | AML · WOD unsupervised moving-object 3D detection (PointPillars) | 3D mAP L1: 9.6 -> 42.1 | Improved |
| replace unsupervised flow with supervised flow in AML | AML · WOD unsupervised moving-object 3D detection (PointPillars) | 3D mAP L1: 42.1 -> 49.9 | Improved |
| add flow-based clustering to position clustering (ablation) | AML · WOD auto label quality (3D mAP L1, labels) | 3D mAP L1: 25.5 -> 30.4 | Improved |
| add rotation and translation initialization with constrained ICP to registration (ablation) | AML · WOD auto label quality (3D mAP L1, labels) | 3D mAP L1: 32.2 -> 36.9 | Improved |
| constrain ICP rotation to z-axis (versus unconstrained) (ablation) | AML · WOD auto label quality (3D mAP L1, labels) | 3D mAP L1: 34.3 -> 36.9 | Improved |
Closest papers by meaning
- Unsupervised 3D Perception with 2D Vision-Language Distillation for Autonomous Driving (ICCV 2023 · similarity 0.91)
- SEPT: Towards Efficient Scene Representation Learning for Motion Prediction (ICLR 2024 · similarity 0.87)
- Prioritizing Perception-Guided Self-Supervision: A New Paradigm for Causal Modeling in End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.87)
- SmartPretrain: Model-Agnostic and Dataset-Agnostic Representation Learning for Motion Prediction (ICLR 2025 · similarity 0.87)
- Enhancing End-to-End Autonomous Driving with Latent World Model (ICLR 2025 · similarity 0.87)
- OpenAD: Open-World Autonomous Driving Benchmark for 3D Object Detection (NeurIPS 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
